Short-Run and Long-run Relationships in the Consumption of Alcohol in the Scandinavian Countries

7/27/2019 Short-Run and Long-run Relationships in the Consumption of Alcohol in the Scandinavian Countries

1/22

ISBN 87-7882-010-3 (trykt)ISBN 87-7882-011-1 (online)

WORKING PAPER 04-14

J an Bentzen and Valdemar Smith

Short-run and long-run relationships in theconsumption of alcohol in the Scandinavian

countries

Department of EconomicsAarhus School of Business


2/22

Short-run and long-run relationships in the consumption

of alcohol in the Scandinavian countries

Jan Bentzen

Valdemar Smith

Department of Economics

Aarhus School of Business

Silkeborgvej 2, PrismetDK-8000 Aarhus C

Denmark

E-mail: [email protected]; [email protected]

Abstract:

Traditionally, the Scandinavian countries have been characterized as spirits and beer consuming

countries and a century ago the historical background was decades of relatively heavy drinking

behaviour with spirit as the preferred beverage. Therefore, it might be expected that alcohol

consumption especially in the last part of the 19th century and the beginning of the 20th

century would behave in a counter-cyclical manner, i.e. heavy drinking during severe

recessions characterized by harsh economic conditions. Using long-run time series data for

alcohol consumption levels in the Scandinavian countries the question of a counter-cyclical or

pro-cyclical behaviour is addressed with the business cycle measured as the GDP and the

empirical findings are that generally, alcohol consumption behaves in a pro-cyclical manner in

the short run, and with no long-run relationship concerning real income.

Keywords: Alcohol consumption, Business cycles, Scandinavia

JEL codes: C22, D12

_____________________________________________________________________________

We thank participants at the VDQS Oenometri XI Conference in Dijon, May 2004, for helpful comments on an

earlier version of the paper.


3/22

1. Introduction

The Scandinavian countries are usually characterized as spirits and beer consuming countries

because these two beverages have been the far most popular alcoholic drinks through centuries.

Until a few decades ago wine was consumed only in very modest quantities in these northern,

climatically rather cold countries. For obvious reasons wine production is not possible or

efficient in the Scandinavian countries and of course this has influenced the drinking behaviour.

But from the beginning of the1960s wine consumption suddenly increased in all of the

Scandinavian countries and during the next three decades wine became a widespread, popular

beverage even with per capita consumption levels (measured in pure alcohol) surpassing

spirits. At present, beer is still the most common alcoholic beverage in all of the three countries,

but in relatively few years from now wine may well be the most popular, alcoholic beverage inone or more of the Scandinavian countries.

The drinking patterns in the Scandinavian countries have evolved or converged towards a

continental behaviour as far as alcohol consumption is concerned. Additionally, in the wine

consuming countries of Southern Europe beer has recently gained popularity and therefore a

much more uniform pattern of alcohol consumption is seen today among the European

countries. However, this historical shift in the drinking behaviour in Scandinavian countries

i.e. wine consumption becoming widespread and probably substituted for spirits and/or beer is

only the latest episode in a long development and history of each countrys alcohol

consumption.

Significant and even dramatic changes in alcohol consumption did in fact take place regularly

during the latest couple of decades in these countries and the aim of this paper is to shed light

on the long-run movements in wine, spirits and beer consumption in the Scandinavian

countries. Using long-run time series data for alcohol consumption levels, i.e. from the middle

of the 19th century, the question of a counter-cyclical or pro-cyclical behaviour of alcohol

consumption is addressed - with the business cycle measured as the real GDP.

2. Long-run trends in alcohol consumption in the Scandinavian countries

In the 19th century considerable amounts of both spirits and beer were consumed even

compared to the present levels of alcohol consumption but there is no long-run tradition for

wine drinking in any of the Scandinavian countries. Until the late 1960s wine was considered

partly as a luxury good, which was regularly consumed by only a small part of the population.

Like other alcoholic beverages wine has usually been heavily taxed and therefore relatively

1


4/22

good statistical sources are available when it comes to alcohol consumption levels. Thus, for

Norway and Sweden per capita consumption levels of the specific beverages are published back

to the 1850s and for Denmark data are available since the beginning of the 1890s with some

exceptions.

Figure 1 shows the long-run development in the per capita alcohol consumption levels in each

of the Scandinavian countries (litres per capita aged fifteen years and above) - the statistical

sources concerning alcohol consumption are presented in more detail in the data appendix. It is

clear that significant changes have taken place in the overall alcohol consumption over the past

100-150 years most dramatic for Denmark. In fact, the Danish per capita alcohol consumption

level was as high as 15 litres in the late 19th century, which is even higher than todays

consumption level. However, from the beginning of the 20th century the alcohol consumption

dropped drastically to a level around 5-6 litres per capita where it stayed until the 1960s. Since

then the Danish alcohol consumption has increased mainly driven by higher consumption of

beer and wine.

Figure 1. Alcohol consumption in the Scandinavian countries 1851-2002 (litres per capita,

100 pct alcohol, 15 years+).

0

2

4

6

8

10

12

14

16

1851 1871 1891 1911 1931 1951 1971 1991

Denmark Norway Sweden

Notes: The data are calculated as litres per capita for the part of the population aged 15 years or above.

Sources: Statistical sources as listed in the data appendix.

Contrary to the situation in many other countries there have been no periods of prohibition in

Denmark and even though the temperance movement existed since the late part of the 19th

2


5/22

century it never had any decisive influence on the Danish alcohol policies. Even though the

temperance movements asked for prohibition in connection with World War I no policy actions

in terms of ban and similar actions were taken. However, significant tax increases on alcohol

were implemented in 1917 and the Danish alcohol taxes were at a relatively high level untilDenmark joined the European Union in 1972. The main reason for the high alcohol taxes and

duties was predominantly fiscal.

The Swedish experience is in accordance with the development in Denmark. Despite rather

strong temperance movements, direct prohibition was not implemented. Instead taxes on

alcohol were increased and in connection with World War I production and sales of alcohol was

monopolised by the state from 1919, and the retail sale of alcohol was rationed by the

introduction of the so-called Bratt System a ration book for each individual regulating the

amounts of alcohol bought. The system was later abandoned in the 1950s.

Two centuries ago Norway experienced significant alcohol problems and per capita alcohol

consumption was as high as 13 litres around 1830-40. Naturally, the temperance movements

gained momentum and in 1919 direct prohibition was decided. However, due to protests from

wine producing countries, the ban on wine ceased in 1922 and so did the ban on strong wine in

1923. Finally, in 1926 sales of strong alcohol was allowed again. At the same time the retail and

wholesale alcohol sales system changed to a system of state monopoly which still exists.

Spirits and beer have always been produced in all of the Scandinavian countries and

consequently, they have been integrated parts of the food and drinking culture for centuries.

Figures 2-4 also show the significant structural shifts with strongly increasing shares for wine

and the marked declines for spirits in Norway and Sweden. Today, beer is in fact the most

popular drink but this might look different in the future if the present trends in beverage shares

continue. Denmark is the first of the Scandinavian countries in the post WWII period to

become a place where wine is regularly consumed and in levels comparable to other European

countries. Being a part of continental Europe Denmark has adhered to relatively more liberal

alcohol policies compared to Norway and Sweden, which may explain the higher Danish

consumption levels. Of course, Denmark lags behind countries in Southern Europe concerning

wine consumption, e.g. Italy with approximately 50 litres of wine per capita and for France the

annual level is 60 litres per capita (direct measures in volumes), cf. also Anderson (2004).

3


6/22

Figure 2. Consumption of alcoholic beverages, Denmark 1893-2002 (litres per capita, 100

pct alcohol, 15 years+).

0

2

4

6

8

10

12

14

16

1893 1903 1913 1923 1933 1943 1953 1963 1973 1983 1993

Total Beer Wine Spirits


Figure 3. Consumption of alcoholic beverages, Norway 1851-2002 (litres per capita, 100 pct

alcohol, 15 years+).

0

1

2

3

4

5

6

7

8

1851 1866 1881 1896 1911 1926 1941 1956 1971 1986 2001



4


7/22

Figure 4. Consumption of alcoholic beverages, Sweden 1861-2002 (litres per capita, 100 pct

alcohol, 15 years+).

0

2

4

6

8

10

12

1861 1876 1891 1906 1921 1936 1951 1966 1981 1996



In spite of the restrictive alcohol policies i.e. regulations and taxation all Scandinavian

countries seem to have entered tracks of strongly increasing wine consumption levels, whereNorway with the most restrictive policy seems to lag most behind. The main reasons for

increases in wine consumption are probably the rising living standards steaming from the strong

growth rates of real incomes and more open economies, i.e. especially the tendencies developed

from the early 1960s - the Scandinavian populations travelling on holidays to Southern Europe -

where wine was also abundantly and cheaply available compared to the conditions in the home

countries! As no grape wine production takes place in the Scandinavian countries consumer

habits or preferences for wine had to be adopted from outside.

The national retail systems

As mentioned above state monopolies in production, trade and sales were established in

Norway and Sweden at the beginning of the 20th century for alcoholic beverages, although there

were some exemptions for beer. Still today, Norway and Sweden have state monopolies,

although they are only in effect regarding retail sales of alcohol (again with some exemptions

for beer). For the production and wholesale of alcohol beer, wine or spirits - a licence is

needed and furthermore, spirits production is a state monopoly in Norway. Denmark is the

exception with none of these regulatory systems. However, all the Scandinavian countries have

5


8/22

used heavy taxation in order to create considerable fiscal revenues and - usually presented as an

important motive - also to curb all kinds of alcohol consumption, cf. Bentzen and Smith (2004).

Sweden has organized the retail system of alcoholic beverages as a state-owned monopolySystembolaget operating through shops (less than 500) or local agents in 575 communities.

The monopoly covers spirits, wine and strong beer. But light beer is available in other shops too

and the Swedish breweries can sell strong beer (above 3.5% alcohol) directly to restaurants. At

the retail level spirits, wine and strong beer may only be sold to persons aged 20 years or more.

Today, there are no quantitative restrictions on the amounts of wine individuals may buy, but as

the Systembolaget is a state monopoly with a policy of supporting a healthy drinking culture

they are not striving to attract customers to the shops.

Since the 1950s, some effort has been devoted to the objective of making customers substitute

wine for spirits because wine is considered healthier relative to spirits and therefore a relatively

faster introduction of wine in the consumption of alcoholic beverages has taken place despite

the otherwise restrictive retail sales system. Until 1994 Systembolaget had an effective

monopoly also in the case of quantities sold to e.g. restaurants, but this part of the monopoly

ceased in 1995 and there are currently 200 licensed, private import companies supplying wines

to Systembolaget, restaurants, hotels, etc.

After Sweden joined the European Union in 1995 problems were expected for the sales

monopoly in the market-oriented community. However, in 1997 the EU Court of Justice ruled

that Systembolaget was not in contradiction with an EU membership because it was created

due to public health considerations. Moreover the Swedish system was not found to

discriminate between foreign and Swedish products. Still a gradual liberalization of the border

trade regulations, i.e. personal imports of (cheaper) alcoholic beverages, has taken place and

more important the monopolies in import/export, wholesales and production (spirits) have

disappeared.

Norway has a system rather similar to the Swedish system and Vinmonopolet was established

in order to control a widespread misuse of spirits aqua vitae. Norway is not a member of the

European Union, but partly due to trade-agreements and the general liberalization of

international trade the state-monopoly was split up in the mid 1990s and, similar to the

changes in Sweden, Vinmonopolet is today only a retail sales monopoly. In total, there is less

than 200 sales outlets, and the low density of alcohol stores combined with the geography of

Norway, i.e. mountains, forests, a lot of snow in the wintertime, etc. makes it difficult for

people living outside urban areas to buy alcohol.

6


9/22

Traditionally, Denmark has adhered to liberal, market-oriented systems. Taxes have been used

primarily for fiscal purposes and except for a minimum age of 18 years for sale at restaurants,

there have generally been no impediments to alcohol consumption. Recently, a limit of 15 yearswas introduced for the sale of alcohol from retail shops. Furthermore, after the Danish

membership of the European Union in 1972, free trade and harmonization of tax systems have

contributed to the continuous increase in the absolute alcohol consumption levels, especially for

wine cf. figure 2 - which shows a significant increase in the consumption of wine from the

early seventies. In fact, a tax cut of 75 per cent on wine was effectuated after 1972, and the tax

rates on beer and spirits were fixed in absolute terms, and again in the 1990s a further lowering

of Danish alcohol taxes was effectuated in order to complete the harmonization of the Danish

tax levels towards other countries in the European Union. Accordingly, it is evident that much

more attention has been paid to fiscal consideration and obligations to the EU-membership than

to health and social arguments in the Danish alcohol policies.

Norway and Sweden both decided not to enter the EU in 1972 and hence, both countries were

able to formulate their own fiscal policy, especially with regard to the taxation of alcoholic

beverages. In Sweden real alcohol prices were relatively constant during the 1960s, but they

have been rising since the beginning of the 1970s. Thus fiscal considerations have been of

significant importance to the Swedish taxation of alcoholic beverages. Norway also decided to

stay outside the EU in 1994, by a referendum, and therefore, Norway has no legal problems

withholding a high level of taxation on alcohol - but there are increasing problems concerning

border trade in relation to both Denmark and Sweden.

3. Alcohol consumption and the business cycle - a short-run relationship

From an economic point of view there most likely will to be a positive relationship between

alcohol consumption and real income, but a counter-cyclical relationship might also be present

if for example recessionary periods are accompanied by more heavy drinking behaviour. In

order to address the question of the relationship between the general business cycle - measured

as real GDP per capita - and the consumption of alcohol, scatter plots of log differences of these

variables are exhibited in the figures 5, 6 and 7 for the respective countries. When using values

in log differences, only short-run relationships are revealed as information about level values

are filtered away via the difference operator.

7


10/22

Figure 5. Scatter plot of real GDP per capita and alcohol consumption per capita (15

years+) for Denmark, 1893-2002 (values in log differences).

-0,25-0,20

-0,15

-0,10

-0,05

0,00

0,05

0,10

0,15

0,20

0,25

-0,20 -0,15 -0,10 -0,05 0,00 0,05 0,10 0,15

GDP/cap.

A

lc./cap.


years+) for Norway, 1865-2002 (values in log differences).

-0,5

-0,4

-0,3

-0,2

-0,1

0,0

0,1

0,2

0,3

-0,15 -0,10 -0,05 0,00 0,05 0,10 0,15

GDP/cap.

A

lc./cap.

Note: A few outliers with extreme values are not included in the graph.

8


11/22


years+) for Sweden, 1861-2002 (values in log differences).

-0,25-0,20

-0,15

-0,10

-0,05

0,00

0,05

0,10

0,15

0,20

0,25

-0,1 -0,05 0 0,05 0,1 0,15

GDP/cap.

A

lc./cap.

Note: A few outliers with extreme values are not included in the graph.

The figures 5-7 reveal that there is no doubt about a pro-cyclical relationship between the

growth rates of GDP and alcohol consumption. This result is also found for sub-samples of the

data - even for data related to the 19th century - and therefore not much evidence is found infavour of a hypothesis of 'heavy drinking in harsh times' as an escape from poverty and severe

social problems. For the specific beverages, i.e. beer, wine and spirits, the cross correlation

coefficients are reported in table 1.

Table 1. Cross correlations between GDP per capita and per capita consumption of beer,

wine and spirits (Log differences).


GDP : Beer 0.461 0.345 0.244

GDP : Wine 0.178 0.406 0.118

GDP : Spirits 0.201 0.154 0.359

Note: The data used when calculating the correlation coefficients are 1893-2002 for Denmark, 1865-2002 for

Norway (the data for GDP only covers this period, but the alcohol consumption data goes back to 1851) and 1861-

2002 for Sweden.

9


12/22

The conclusion from table 1 is in accordance with the plots in the figures above as all

correlations are found to be positive. Hence, in the short run there is a pro-cyclical pattern for

the overall consumption of alcohol as well as for the specific alcoholic beverages. In the longrun there is less doubt about substitution processes taking place between beer, wine and spirits

as exhibited in the graphs in figures 2-4. Looking at these data covering more than a century,

the general pattern for all the Scandinavian countries is a steady decline in the consumption of

spirits in favour of increasing consumption of beer and wine - where the consumption levels of

the latter has increased tremendously since the 1960s and probably will be the most important

alcoholic beverage in the future evaluated from the present trends in consumer behaviour.

From an economic point of view alcohol consumption is expected to be influenced by real

income and real prices of the specific alcoholic beverages - as well as other variables, of course.

From the presented graphs and correlation analysis there seems to be evidence of a positive

income effect, but data for prices of the alcoholic beverages are not available for the long time

span covered by the consumption and income data already exhibited1. Therefore, in order to

obtain an estimate of the (short-run) income effect in alcohol consumption an autoregressive

distributed lag model (ARDL) is applied to the data:

0 1ln ln ln

p q

t i t i i t i iA Y

i tA = = = + + +

(1)

First differences of income (Y) and alcohol consumption (A) measured in per capita terms are

included in the RHS of (1) with a suitable number of lags following the general-to-specific

estimation technique. The income variable is assumed exogenous and the parameter estimates

are reported in table 2.

1

See Bentzen et al. (1997, 2004) and Milhj (1993, 1995) for prices concerning alcoholic beverages and priceelasticities.

10


13/22

Table 2. Parameter estimates of the ARDL models.


-0.019* (-2.78) -0.030* (-2.68) -0.034* (-3.07)

0 0.806* (5.37) 1.480* (5.06) 1.444* (5.46)

1 - - - - 0.321* (4.87)

2 0.252* (3.01) - - -0.512* (-7.80)

2

R 0.25 0.15 0.44

DW 1.82 2.20 1.93

LM(2) 1.29 (0.52) 2.08 (0.35) 3.57 (0.17)

LMARCH(2) 0.35 (0.84) 5.41 (0.07) 39.72 (0.00)

Note: t-values of the parameter estimates in parenthesis and with a * indicating significance at least at the 5 per

cent level. LM(2) and LMARCH(2) are Lagrange Multiplier tests for second order residual autocorrelation and

second order autoregressive conditional heteroscedasticity, respectively, withp-values in parenthesis.

The resulting parameter estimates are found to be very significant, although the models for

Norway and Sweden still suffer from ARCH errors, which will remain even when extending the

number of lags. Assuming the ARDL models are reasonably well representing short-run

relationships in alcohol consumption - in spite of the missing price variables, which may cause

bias in the parameter estimates - the best available estimates of the income elasticities are the

0-parameter estimates. Thus, the short-run income elasticities are relatively high, especially for

Norway and Sweden, and even if the magnitudes are somewhat exaggerated due to missing

variables, the conclusion is still a strongly pro-cyclical alcohol consumption pattern.

In the long run no evidence for at positive relationship between income and alcohol

consumption levels is expected - for obvious reasons as there will be a natural limit to the

human intake of alcohol. Evaluated from table 2 the long-run elasticity between the growth

rates of income and alcohol consumption will be zero for all three countries. As first

differences of the variables are estimated in (1), only relationships between growth rates can be

evaluated - not level values and the parameter estimates in table 2 all imply a zero long-run

relationship between alcohol and real income. Consequently, the next question to address in the

following part 4 will be whether there is some kind of 'mean reversion' in the long-run level of

alcohol consumption for the Scandinavian countries.

11


14/22

4. Alcohol consumption in the long run: a mean reversion process or a

random walk?

When evaluating the data for alcohol consumption from figure 1 these variables seem very

much to exhibit random walk behaviour, i.e. be non-stationary I(1)-variables. Testing for unit

root behaviour by the Dickey-Fuller test statistic is reported in table 3.

Table 3. Unit root test of alcohol consumption levels (log values).


DF/ADF -1.068{0} -2.212{0} -2.342{2}

N 109 151 139

Note: N is the number of observations, and the number of lags in the augmented DF-test indicated by the {}-

parenthesis.

The conclusion from the DF/ADF-tests is in all three cases non-stationarity as the critical value

is -2.88 at the 5% level of significance according to MacKinnon (1991). When taking a closer

look at figure 1 an alternative interpretation of the unit root hypothesis - especially for Norway

and Sweden - might be an asymmetric adjustment process towards a long-run level of alcohol

consumption that is non-zero. Such an alternative hypothesis to the random walk behaviour as

tested for in table 3 might be the so-called threshold autoregressive model (TAR-model) as

presented in Enders and Granger (1998), see also Enders and Siklos (2001). In the present case

of alcohol consumption levels the most obvious version of these models will be the TAR model

with a linear attractor - the latter being the average long-run level of alcohol consumption. The

threshold model (TAR) with a linear attractor is defined by (2) and (3), see Enders and Granger

for further details:

( ) ( ) ( )1 1 2 11

ln ln 1 ln lnp

t t t t t i t i

i

A I A A I A A A t =

= + + + (2)

where the linear attractor (A )is the average value ofAtand t is a white noise disturbance term.

The indicator function,It, is defined as:

( )( )1 ln 1

0 ln1

{t

At

A At

I

Date post:	02-Apr-2018
Category:	Documents
Upload:	marhelun
View:	218 times
Download:	0 times

Short-Run and Long-run Relationships in the Consumption of Alcohol in the Scandinavian Countries

Documents